Mistral Small 4 vs Mistral Large 3: entry-tier utility vs flagship reasoning
Mistral Small 4 is over 3x cheaper and perfect for fast, high-volume classification or summarizing tasks. Mistral Large 3 is excellent for complex reasoning and deep multi-lingual synthesis.
Utility-scale model versus flagship Europe-hosted logic.
Choosing between Mistral's lightweight utility model and its premium large frontier engine involves analyzing your API traffic volumes.
By TechCompare · Updated
Cost Comparison
Based on 100,000 input tokens (50% cached), 5,000 output tokens, and 100 requests.
Side-by-side specs
| Spec | Mistral Small 4 | Mistral Large 3 |
|---|---|---|
| Input Cost (per M) | $0.15 (better on this spec) | $0.50 |
| Output Cost (per M) | $0.60 (better on this spec) | $2.00 |
| Cached Input (per M) | $0.015 (better on this spec) | $0.05 |
| Batch Discount | 50% | 50% |
How they differ
Mistral Small 4 is priced at $0.15 per million input tokens and $0.60 per million output tokens. Mistral Large 3 runs at $0.50 per million input tokens and $1.50 per million output tokens. Both offer a 90% cache discount and 50% batch discount.
Verdict
Small 4 wins every base row: $0.15/M input against $0.50 and $0.60/M output against $2.00, both about 3x cheaper. The 90% cache discount lands at $0.015/M versus $0.05/M, a 3.3x gap that mirrors the base ranking. Both apply the same 50% batch discount, so the ratio holds in every mode. Large 3 earns its premium only on flagship multi-lingual reasoning and code-quality work Small 4 cannot pull off.
Which should you pick?
Choose Mistral Small 4
High-frequency summarization, simple chatbot agents, and low-latency utilities.
Choose Mistral Large 3
Sovereign European logic setups, advanced code writing, and translation pipelines.